The meta, under glass.
Who wins. What gets bought. Where the ranks land. Aggregate observations from the community API — not a census of Deadlock, and not Valve’s balance plans.
Current sample: 2026-10-10 – 2026-10-16 UTC. Normal mode, ranked + unranked matches. Source cache: 15 minutes; reload for updates. Upstream item results may be cached for 6 hours.
Hero win-rate spread
All returned heroes, ordered by win rate. Each dot is wins ÷ hero match observations; a match can contribute to several heroes. Axis: 0–100%.
Source query ↗ · Retrieved 2026-10-17T08:13:54.990Z
Nerf-candidate watchlist
A ranking of above-50% results, not a probability of a nerf. Score = max(0, Wilson lower bound − 50%) in percentage points; z = 1.96, minimum 1,000 observations. Shared matches and selection bias mean the bound is a ranking device, not a guarantee.
Source query ↗ · Retrieved 2026-10-17T08:13:54.990Z
No ban-rate weights, undocumented coefficients, or claimed predictive accuracy.
The most-observed items
Top 20 by source purchase observations. Bars show buyer win rate, not the benefit caused by buying an item. Winning players can afford different builds; multiple items can occur in one match.
Source query ↗ · Retrieved 2026-10-17T08:13:55.001Z
Item win-rate changes
Before: 2026-10-03 – 2026-10-09 UTC.
After: 2026-10-10 – 2026-10-16 UTC.
Largest 20 absolute changes among items with ≥1,000 observations in each period. Equal seven-day windows, not a patch boundary or causal patch effect. Positive means a higher observed win rate.
Source query ↗ · Retrieved 2026-10-17T08:13:55.001Z
Source query ↗ · Retrieved 2026-10-17T08:13:55.008Z
Badge distribution · MMR proxy
Unique players by the rank reported at their latest ranked match within the window. Badge codes are tier × 10 + subtier. This is a badge distribution, not raw hidden MMR. Match counts use average match badge and are a different measure.
Source query ↗ · Retrieved 2026-10-17T08:13:55.015Z
Coverage, not a time machine.
Storage: Redis. 4 successful collection snapshots retained (maximum seven days).
Observed coverage: 2026-10-17T06:30:29.596Z to 2026-10-17T07:45:29.798Z. Gaps are not interpolated. These snapshots record collection coverage, not historical player counts.
Collection ledger and methodology
| Observed UTC | Hero rows | Item rows |
|---|---|---|
| 2026-10-17T06:30:29.596Z | 43 | 156 |
| 2026-10-17T06:45:29.798Z | 43 | 156 |
| 2026-10-17T07:15:29.779Z | 43 | 156 |
| 2026-10-17T07:45:29.798Z | 43 | 156 |
Temporal windows end at the last completed UTC day. Names come from the verified v1 asset catalogs; catalog failures retain numeric IDs. No official art is loaded. Source population coverage and ingestion delay are not known. Percentage displays are rounded; exact numerators and denominators remain available on every row.
Source query ↗ · Retrieved 2026-10-17T08:13:55.020Z
Source query ↗ · Retrieved 2026-10-17T08:13:55.087Z
Back to the patch tracker · Data by deadlock-api.com. Fan project. Not affiliated with Valve Corporation.